Proper design wave height estimation is very important for the design of ocean and coastal engineering, for example, the structural design of wharfs and breakwaters and the construction of ocean platforms. In recent years, performing the design wave height estimation by a Peak Over Threshold (POT) method and in combination with a Generalized Pareto Distribution (GPD) method has been widely applied in the ocean and coastal engineering. The POT method is a sampling method in which independent large values exceeding a certain fixed value (threshold) are selected as samples. The GPD method is a probability distribution model which can perform design wave height estimation, in different return periods, on the peak over threshold extracted by the POT method.
The key to perform design wave eight estimation by the POT/GPD method is the selection of thresholds. The conventional threshold selection method is a GPD parameter graphical method, that is, a lower boundary value within a stable GPD parameter interval is selected as a threshold. The conventional method has the following disadvantages: the influence from human factors is strong, a higher degree of professional knowledge is required for subjective judgment, the fluctuation in GPD parameters is large, and the stable threshold value is not obvious and not unique, and the conventional method cannot be applied regionally. As an indirect threshold selection method, a threshold is determined by the GPD parameter graphical method and the design wave height estimation is then performed. In this indirect threshold selection method, the sample stability is obtained by determining the stability of GPD parameter, and a reasonable threshold corresponding to the stable design wave height is then obtained.